The July newsletter highlights events, research initiatives, and collaborations happening as part of the AI4AGRI project. Key upcoming events include a visit by Șerban Oprișescu, a lecturer from Transilvania University of Brașov, to Université Toulouse 3 Paul Sabatier in September 2024, where he will deliver a talk on hyperspectral satellite imaging for agricultural applications. Additionally, two PhD students from the university of Rome Tor Vergata will defend their thesis. Another significant event is the presentation by Makoto P. Kato on July 17, 2024, discussing the integration of numerical data in search engines to improve the reliability of information retrieval, addressing challenges related to misinformation on the web.
Past events include a presentation from Kamal Marandinskiy and Professor Mihai Ivanovici on crop identification, focusing on potato field identification with promising results. Mohammad El Sakka delivered a talk at IRIT in Toulouse on the role of AI in smart agriculture, emphasizing how convolutional neural networks can analyze satellite, drone, and ground vehicle imagery to enhance farming practices.
The newsletter concludes by announcing the next AI4AGRI monthly meeting, scheduled for July 16, 2024, where participants will discuss the latest developments and collaborative efforts within the project.
Member Highlight
Șerban OPRIȘESCU
Lecturer – UNITBV, Brasov (Romania)
Serban Oprisescu holds a PhD in electronics and telecommunications from Politehnica University of Bucharest, Romania in 2007. He is currently a lecturer with Transilvania University of Brașov, Romania. He holds a B.S. degree in electrical engineering and computer science (2002) and an M.S. degree in bio-medical engineering (2003), both from University Politehnica of Bucharest. His scientific interests cover image and video processing, ToF cameras, biomedical image processing and analysis and computer science.
News
Website redesign
The UT3 team has undertaken a comprehensive redesign of the AI4AGRI website, implementing several key changes to enhance user experience and access to information. The resources page has been reorganized into two distinct pages: Publications and Training Material. The news page has been split into a list of articles, each dedicated to describing an upcoming or past event, facilitating easier navigation. Additionally, a newsletter has been created to keep stakeholders informed about project developments.
The front page has been restructured to prominently feature the latest posts, upcoming events, and a carousel displaying the latest articles from each post category. This reorganization ensures that visitors can quickly find the most recent and relevant information. Furthermore, a partners page has been created, providing detailed descriptions of the research labs involved in the AI4AGRI project, highlighting collaborative efforts and contributions.
AI4Farms Amorce project
The AI4Farms project has secured funding through the AAP AMORCE, aimed at consolidating and strengthening the consortium while refining the research roadmap for the upcoming HE-MSCA-DN proposal. Following a workshop supported by the CNRS and presented by UT3 and CNRS representatives in December 2023, which highlighted HORIZON EUROPE MSCA opportunities, the project will facilitate collaboration through meetings and workshops. The initial meeting took place in Brasov in May during the AI4AGRI summer school, and subsequent online meetings are scheduled for July, with the proposal slated for completion in November. The allocated 5,000 euros will fund these preparatory activities, ensuring the project’s readiness for the MSCA Doctoral Network submission.
Upcoming Events
Șerban OPRIȘESCU visit in Université Toulouse 3 Paul Sabatier
🧑 Șerban OPRIȘESCU
📅 Tue. 24 – Sat. 28 September 2024
📍IRIT – Toulouse (France)
Image analysis (features, segmentation) for applications in agriculture
🧑 Șerban OPRIȘESCU
📅 Wed. 25 September 2024 – 11:00 (Fr time) | 12:00 (Ro time)
📍Salle des thèses, IRIT – Toulouse (France)
https://univ-tlse2.zoom.us/j/98366074118?pwd=VFVOdVdacTdtSzVNelhvei9Wa3FOQT09
Hyperspectral satellite imaging offers high spectral resolution images of a scene in hundreds of narrow spectral bands. This remote sensing technique proves to be very useful in many Earth Observation applications such as agriculture crop health assessment, land cover mapping and other tasks. The first part of the talk, after a brief presentation of the author’s research activity, shows some results on the semi-automatic estimation of the Shannon-Weaver biodiversity index in hyperspectral images. The purpose is the qualitative analysis of grassland areas. After an image segmentation based on histogram thresholding of spectral angle mapper (SAM) values, we compute the entropy for the pixels belonging to the segmented grassland areas using a clustering approach. The second part of the talk presents a weakly supervised framework for early identification of autumn wheat in PRISMA images.
Advanced Methodologies for Atmospheric Remote Sensing using Artificial Intelligence (PhD defense)
🧑 Ilaria Petracca
📅 Mon. 26 July 2024
📍University of Rome Tor Vergata, Rome (Italy)
The study of atmospheric parameters is of primary importance for a deeper understanding of the complex phenomena occurring in the atmosphere, which are closely related to the environmental impact of climate change. In this work, advanced remote sensing applications by means of neural networks using UAS (Unmanned Aerial System) and novel space-borne sensors are described, from UAS-based observations for BRDF (Bidirectional Reflectance Distribution Function) retrieval and modeling to the monitoring of atmospheric parameters with a focus on precipitation retrieval in tropical cyclones and ash cloud detection in volcanic eruptions.
Application of Artificial Intelligence algorithms for satellite on-board image processing (PhD defense)
🧑 Giorgia Guerrisi
📅 Mon. 26 July 2024
📍University of Rome Tor Vergata, Rome (Italy)
Earth observation satellites collect huge amounts of data, and new methods are required to process them efficiently. Onboard artificial intelligence (AI), particularly Deep Learning, represents a promising solution, but constraints in the space environment limit the processing power available. This thesis explores the use of optimized AI models for on-board processing tasks, even in a real edge context, including change detection and image compression, in order to successfully transmit valuable information and reduce the data volume to the ground.
Matching Texts with Numerical Data for Evidence-based Information Retrieval
🧑 Makoto P. Kato,
📅 Wed. 17 July 2024 – 15:00 (Fr time)
📍Salle des thèses, IRIT – Toulouse (France)
https://univ-tlse2.zoom.us/j/98366074118?pwd=VFVOdVdacTdtSzVNelhvei9Wa3FOQT09
We are now facing the problem of misinformation and disinformation on the Web, and search engines are struggling to retrieve reliable information from a vast amount of Web data. One of the possible solutions to this problem is to find reliable evidence supporting a claim on the Web. But what is “reliable evidence”? They can include authorities’ opinions, scientific papers, or wisdom of crowds. However, they are also sometimes subjective as they are outcomes produced by people. This talk discusses some approaches incorporating another type of evidence that is very objective — numerical data — for reliable information access. This talk includes the following topics:(1) Entity retrieval based on numerical attributes, (2) Data search, and (3) Data summarization.
Past Events
DoCO 2024 présentation: Crop Identification Using Sentinel 1 and Sentinel 2 Data
🧑 Kamal Marandinskiy
📅 Wed. 26 June 2024 – 15:30 (Ro time) | 14:30 (Fr time)
📍41A, Iuliu Maniu St., Brașov (Romania)
The research conducted by Kamal Marandskiy and Professor Mihai Ivanovici at Transilvania University of Brașov introduces a method for potato field identification using fused data from Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 Optical Imaging. The approach
relies on ridging patterns during potato seeding, that can be used to identify the potato fields.
Dataset of over 5,000 labeled vectors was used to train and test fully connected artificial neural network with two hidden layers. Ground truth data provided by the National Institute of Research & Development for Potato and Sugar Beet in Brașov. The results demonstrated a
high accuracy in distinguishing potato from non-potato fields, with test accuracy of 93.03%, showing potential of combining SAR and optical imaging to identify crop fields before the growth stage.
AI in Smart Agriculture
🧑 Mohammad El Sakka
📅 Wed. 26 June 2024 – 15:30 (Ro time) | 14:30 (Fr time)
📍Salle des thèses, IRIT – Toulouse (France)
https://univ-tlse2.zoom.us/j/98366074118?pwd=VFVOdVdacTdtSzVNelhvei9Wa3FOQT09
Agriculture is shifting toward smart farming, driven by advanced technologies like Artificial Intelligence. Convolutional Neural Networks act as the “eyes” of AI, allowing it to play a crucial role in this transformation by analyzing images from different sources, such as satellites, drones, and ground vehicles. In the upcoming presentation, Mohammad El Sakka will discuss in more detail how AI helps agriculture, from understanding the needs of plants in agricultural lands to predicting the yield of a field and more.
Upcoming meetings
AI4AGRI Monthly meeting
📅 Tue. 16 July – 14:00 (Fr time) | 15:00 (Ro time)
https://univ-tlse2.zoom.us/j/98366074118?pwd=VFVOdVdacTdtSzVNelhvei9Wa3FOQT09
Latest AI in Agriculture news
Artificial intelligence in the agri-food sector
European Parliamentary Research Service
March 2023, 92 pages
https://www.europarl.europa.eu/RegData/etudes/STUD/2023/734711/EPRS_STU(2023)734711_EN.pdf
The document outlines the objectives of incorporating artificial intelligence (AI) in the agri-food sector, highlighting the need for comprehensive analysis and policy development to maximize the benefits of AI while mitigating potential risks.
The AI4AGRI project received funding from the European Union’s Horizon Europe research and innovation programme under the grant agreement no. 101079136.
Publishing managers: J. Mothe & S. Molina, UT3 & UT2, IRIT, France